Video Browsing Effectiveness in Information Seeking Tasks

5. Conclusions

Based on the results of this experiment, the slide show (SS) display was consistently perceived by users to be less useful and more confusing than the storyboard (SB) interface, in spite of the lack of statistically significant differences in task performance. It may be concluded from these results that user preference for interface type is not greatly affected by the type of task to be achieved. Thus, the theory proposed in the Introduction, that satisfaction with a video browsing interface depends on successful task achievement, may need to be reexamined.

Contrary to the original hypothesis, subjective satisfaction is not dependent on successful task performance but rather on perceived "comfort" with the interface. Users found that the rapid flipping of images was more distracting or disorienting for them than any positive performance gains.

One reason may be that users are perceiving small glimpses of the images in the SS display and storing information in preattentive memory, where the information is not reaching consciousness. Since the still images are being shown at 3 key frames per second, each image is on the screen for about 333 ms. On average, object recognition under controlled conditions takes about 100 ms (Potter, 1976). Thus 330 ms does not allow users to recognize many objects before the image changes. Also, as display flips through the images, users will likely need to reorient themselves each time. It is possible that 3 reorientations each second is too disruptive/disorienting for users to feel comfortable.

Recognition and information storage must occur rapidly as the subject immediately views the next still image. A possible mechanism is that objects are stored and quickly recalled during the task completion phase. Another possible mechanism is that users create schemas or "story lines" in their minds or have reasonable guesses as to "what is happening" in the video based on context and order of the stills.

The results presented here cannot rule out either mechanism. During the object recognition (OR) task, users are able to distinguish distractors from actual objects that appear in the video, despite the fact that the distractor objects were selected to support a common theme (e.g., bow and arrow in a clip about Cherokee tribes). On the other hand, subjects were able to select the best concepts or themes as provided by a "control group" that had an opportunity to watch the actual video clip with the audio soundtrack despite the poor user satisfaction scores.

Another reason for the significant user dissatisfaction with the SS interface could be the lack of user control for key frame rate and/or direction of play. Unlike the SB design, which is static and allowed users to view and review the images at their own pace, the SS interface "blasts" the images at users at a predefined rate (3 key frames per second), direction of play (forward), and automatically loops. Perhaps implementing different widgets such as buttons, scrollbars, or Velocity Sliders to allow users to control the SS inteface would improve subjective satisfaction ratings. [Shirinian et al. (1997) found that such widgets had an effect on user satisfaction for static overviews. It would be interesting to test their effect on dynamic surrogates.]

As some of the user comments indicated, subjects felt overwhelmed and pressured with the SS design. Also, SS did not allow users to concentrate on a particular image. Instead, they had to wait for the image to cycle around. There was clearly not enough user control over searching and scanning the key frames. This violation of one of Shneiderman's (1998) Golden Rules of Interface Design, "Support internal locus of control," most likely contributed to the dissatisfaction with the SS approach.

Thus, although there are no deleterious performance effects in using the SS interface for the tasks provided, it is clearly not likely to be a good surrogate for the video browsing tasks tested in this experiment. It is recommended that further studies be done on dynamic video representations before they are incorporated into actual systems. User satisfaction results indicate that subjects who react negatively to the SS design and performance measures indicated no advantages over SB.

Although no statistically significant differences were found between the task and interface design variables in this experiment, there are many unanswered questions remaining. Some of these relate to improvements in methodology while others relate to further research in the area.

Methodological Improvements

Increase the sample size (e.g., n = 60?)
Improve user tasks so that even small effects could be detected
Experiment with slower speeds for the SS interface (e.g., 1 key frame per 2 seconds) to determine levels of acceptable subjective satisfaction
Provide user control over the interface (e.g., VCR-like buttons for frame rate speed and direction of play

Future Research Questions

What are other types of user needs could be facilitated by a video browsing interface?
How do user search strategies change depending on the type of surrogate available to them?
Are certain types of surrogates more appropriate for specific user characteristics in addition to user tasks?
What are other types of dynamic video surrogates that preserve the temporal nature of moving images while providing a rapid overview or preview of videos?
Are dynamic surrogates effect ways to abstract video?
Video surrogates have been implmented in the Baltimore Learning Community project (Ding and Marchionini, 1998). How effective have the teachers using these surrogates found them to be for information seeking?

Continue